DESE: estimating driver tissues by selective expression of genes associated with complex diseases or traits

Abstract

The driver tissues or cell types in which susceptibility genes initiate diseases remain elusive. We develop a unified framework to detect the causal tissues of complex diseases or traits according to selective expression of disease-associated genes in genome-wide association studies (GWASs). This framework consists of three components which run iteratively to produce a converged prioritization list of driver tissues. Additionally, this framework also outputs a list of prioritized genes as a byproduct. We apply the framework to six representative complex diseases or traits with GWAS summary statistics, which leads to the estimation of the lung as an associated tissue of rheumatoid arthritis.Electronic supplementary materialThe online version of this article (10.1186/s13059-019-1801-5) contains supplementary material, which is available to authorized users.

Systematic comparative analysis of Mendelian randomization methods for inferring causal genes of complex phenotypes and the application to psychiatric diseases

“…Many methods are being developed to link gene expression to phenotypes. Others' and our recent studies also showed that phenotype-associated genes tend to have selected expression in phenotype related tissues or cell-types; and gene's selective expression was convincing evidence for prioritizing phenotype-associated genes and related tissues 7,8 . Some methods have used expression quantitative trait loci (eQTLs) analysis as an auxiliary instrument to interpret GWAS signals because GWAS hits are found to be enriched with eQTL 9 and being eQTLs are prerequisite for the loci to regulate a phenotype by gene expression 10 9 .…”

“…In addition, gene expressions are variable in different brain regions and development stages. Our previous study showed that multiple brain regions were highly related to schizophrenia and bipolar disorder besides the frontal cortex 7 . Due to limited sources of expression data, we only used the dorso-lateral prefrontal cortex regions.…”

A global overview of single-cell type selectivity and pleiotropy in complex diseases and traits

“…However, probably due to lack of effective methods and data, deciphering cellular spectrum and pathological mechanisms of many complex phenotypes (diseases and traits) is lagging far behind their genome-wide association studies even for well-studied diseases, including schizophrenia 4 , arthritis 5 and type 2 diabetes 6 . Recently, based on genes' selective expression in bulk cells, summary statistics of genome-wide association studies (GWASs) of the complex phenotypes were utilized to explore the tissue specificity of the corresponding phenotypes 7,8 . These studies suggested that genes' selective expression leveraged with GWAS signals may be useful for inferring cellular spectrum of many complex phenotypes.…”

“…The underlying assumption was that the disease susceptibility genes tend to selectively expressed in disease relevant cell-types. The iterative prioritization procedure was extended from our recent method 8 . The major extension was that a more robust conditional gene-based association test was proposed for the procedure (See details in the method section).…”

“…Next, we calculated the selective expression profiles of the clusters by a robust-regression z-score method (REZ) that we previously proposed 8 . To check whether the usage of selective expression can improve discrimination between cell clusters, we investigated the Pearson correlation coefficient of inferred cell types according to original expression and selective expression respectively ( Fig.…”